The canonical model of the dentate gyrus (DG) of the hippocampus suggests the DG performs pattern separation, orthogonalizing similar input patterns [1, 2] aided by an approximate 5-fold expansion of the cell population relative to its entorhinal cortex inputs [3]. However, more recent experimental results challenge this standard model, suggesting the DG also supports the precise binding of objects and events to space and the integration of information across episodes [4]. Very recent studies attribute pattern separation and pattern integration to anatomically distinct parts of the DG, the suprapyramidal blade and the infrapyramidal blade respectively [5, 6]. Several models have investigated pattern separation [2], or the role of adult neurogenesis in the DG [7]. However, none have considered the role of the distinct DG blades. Here we propose the first computational model that investigates this distinction. In line with recent experimental work [6, 8] we hypothesise that the suprapyramidal blade contributes to the storage of distinct episodic memories (via pattern separation and one-shot learning). In contrast, the infrapyramidal blade integrates information across episodes (learning at a slower rate) to form generalised expectations across episodes, eventually forming a cognitive map. In the model, both new and old episodes can be compared to these learned expectations (here, expectations of positions of objects relative to oneself in a spatial layout). This comparison allows for the calculation of a prediction error, which can drive the storage of poorly predicted memories and the forgetting of well-predicted memories, thus allowing the hippocampal system to free up neuronal resources. This allows the model to iteratively build a spatial cognitive map for a familiar environment on which predictions can be generated by short-scale look-ahead.